SWASD: Sliced-Wasserstein Automated Stationarity Detection

SWASD (Sliced-Wasserstein Automated Stationarity Detection) is a Python package that provides an automated framework for detecting convergence of Markov chains to their stationary distributions. SWASD is applicable to both Markov Chain Monte Carlo (MCMC) and fixed-learning-rate stochastic optimization (FLSO) algorithms.

Installation

From PyPI (once published)

pip install swasd

From source

git clone https://github.com/Manushi22/swasd.git
cd swasd
pip install -e .

Dependencies

SWASD requires Python 3.10+ and the following packages:

• numpy >= 2.0
• scipy >= 1.10
• matplotlib >= 3.7
• tqdm >= 4.66
• arviz >= 0.16
• pystan >= 3.7
• POT >= 0.9.3 (Python Optimal Transport)
• xarray >= 2023.7.0

Examples

See the notebook/ directory for detailed example demonstrating usage of SWASD using FLSO updates.